Narrow Results

In system engineering a practical and difficult problem is the identification of those components that mostly influence the system behaviour with respect to safety and reliability. In this regard, the information provided by importance measures gives useful insights for the safe and efficient operation of the system.

Importance measures have received attention since the early times of risk and reliability analysis and have been defined in several ways for binary components in binary systems. Yet, there are applications for which a multi-state modelling of components and systems is required. This is the case of manufacturing production lines and power generation systems in which the components and the overall systems can operate at different, discrete, levels of performance so that the output provided is a fraction of the nominal design capacity. A similar situation occurs in transportation systems. In the limit, there is the case of systems whose performances are characterized in terms of an infinite set of continuous states. These are, for example, the passive safety systems whose utilization is becoming more and more significant in the advanced and innovative concepts of nuclear reactor design.

In this paper, the definitions of the most frequently used classical importance measures are generalized to multi-state systems constituted by multi-state components. The extensions introduced characterize the importance of a component achieving a given level of performance with respect to the overall multi-state system unavailability and performance. The informative content provided by the introduced measures is illustrated on a simple multi-state system.

In this paper, we consider the component importance analysis of coherent systems subject to common-cause failures. The purpose of component importance analysis is to obtain information regarding a component's contribution or importance to the system reliability. The results from the component importance analysis are key contributors to the system design, tuning, and maintenance activities. Various measures have been proposed for the component importance analysis, but little work has been done to compare their performance. In this research, we investigate and compare a set of nine existing importance measures and select the most informative and appropriate one for guiding the system maintenance activity. An important concern in the traditional fault tree reliability analysis, common-cause failures, is also considered in the component importance analysis. An example is designed and analyzed to show the selection process and to illustrate our efficient method for considering the effects of common-cause failures in the component importance analysis.

Importance measures (IMs) are used for risk-informed decision making in system operations, safety, and maintenance. Traditionally, they are computed within fault tree (FT) analysis. Although FT analysis is a powerful tool to study the reliability and structural characteristics of systems, Bayesian networks (BNs) have shown explicit advantages in modeling and analytical capabilities. In this paper, the traditional definitions of IMs are extended to BNs in order to have more capability in terms of system risk modeling and analysis. Implementation results on a case study illustrate the capability of finding the most important components in a system.